[Ffmpeg-devel-irc] ffmpeg.log.20171104

burek burek021 at gmail.com
Sun Nov 5 03:05:01 EET 2017


[00:07:48 CET] <josefig_> hi, i compiled successfully using the guide for centos: https://trac.ffmpeg.org/wiki/CompilationGuide/Centos now how can I move it on /usr/bin i mean the executables in order to make it visible to anyone
[00:08:44 CET] <teratorn> josefig_: mv?
[00:08:56 CET] <josefig_> teratorn: just moving the binary should be ok?
[00:09:03 CET] <teratorn> josefig_: or 'make install' if you configured properly
[00:09:43 CET] <teratorn> josefig_: it should, but you may need to mv shared libs too
[00:10:00 CET] <teratorn> you should use make install or compile statically
[00:10:14 CET] <josefig_> teratorn: i set it up using the $HOME/bin
[00:10:34 CET] <josefig_> teratorn: ok, i followed this guide but i dont know if that is compiled statically
[00:10:36 CET] <josefig_> :P
[00:12:53 CET] <josefig_> teratorn: ok moved worked :)
[00:12:55 CET] <teratorn> josefig_: it looks like they may be static, if you do `ldd ./ffmpeg` where you compiled the binary do you see it linked against libavcodec etc?
[00:12:57 CET] <josefig_> mv *
[00:13:04 CET] <Johnjay> if I tell a program like vlc to listen on port 554 with rtsp should I see that output in netstat?
[00:13:16 CET] <teratorn> Johnjay: with the right options yes
[00:13:29 CET] <teratorn> Johnjay: and uh maybe only if you are root? its a privledged port
[00:13:34 CET] <Johnjay> i'm struggling like hell to get vlc to play something from rtsp mode with vlm. but i'm failing hard
[00:13:44 CET] <Johnjay> I'm on windows. Should I use port 5554?
[00:13:52 CET] <teratorn> oh windows doesn't have priv ports
[00:14:56 CET] <Johnjay> I see a bunch of threads listening on random ports like 58384
[00:15:07 CET] <Johnjay> but nothing for port 554 or 5554 or whatever
[00:16:22 CET] <Johnjay> I'm giving it the option --rtsp-host 0.0.0.0:554 (or 5554)
[00:16:24 CET] <Johnjay> but so far nothing
[00:17:46 CET] <Johnjay> it says to check the log for details but what log??
[00:22:23 CET] <teratorn> Johnjay: use Process Explorer to see what ports its actually using
[00:22:32 CET] <teratorn> Johnjay: no idea what netstat on windows does
[00:23:55 CET] <Johnjay> netstat works
[00:24:13 CET] <Johnjay> as far as I can tell I'm telling vlc to do rtsp broadcast but it's not doing it. i dont' see the port
[00:25:23 CET] <Johnjay> idk anything.
[00:25:37 CET] <Johnjay> been grinding away at this for 2 hours now and no answer yet
[00:25:55 CET] <Johnjay> using --rtsp-port option doesn't seem to work either
[00:30:04 CET] <DocHopper> alexpigment: I don't have enough bandwidth to send the file I'm generating OTA, so I'll have to limit my bitrate, maybe a higher -crf will be a good thing.
[00:32:04 CET] <Johnjay> heh I found a command that works. dunno how exactly though
[00:36:17 CET] <Johnjay> always a miserable affair when trying to work with vlc for anything
[00:36:45 CET] <hiihiii> hello
[00:37:23 CET] <hiihiii> I'm trying to normalize audio but in a fashion that brings the quit portion closer to the louder ones
[00:38:06 CET] <hiihiii> peak normalization not suited for the inputs I have
[00:38:27 CET] <teratorn> you mean you want to leave the peaks alone?
[00:39:13 CET] <hiihiii> yes
[00:39:18 CET] <hiihiii> something like http://www.zittware.com/Products/CDMaster32/Tutorials/images/Vlevel_compare.png
[00:39:58 CET] <hiihiii> currently I'm just doing volume=-7.0dB
[00:41:06 CET] <teratorn> you could do a normal normalization pass, then increase volume
[00:41:20 CET] <teratorn> otherwise you need a specific filter that does that..
[00:42:37 CET] <teratorn> hiihiii: this appears to do it with the -m option, https://github.com/slhck/ffmpeg-normalize
[00:44:19 CET] <Johnjay> here's the command that seemed to work. lol.
[00:44:21 CET] <Johnjay> https://pastebin.com/Rw5k8NCL
[00:47:39 CET] <Johnjay> now trying with an mp3 file I get error unsupported codec with rtp mp3
[01:01:50 CET] <hiihiii> teratorn, thx I need ffmpeg. my inputs are over a server
[01:02:27 CET] <hiihiii> I'll get a build with loudnorm and try with that
[01:04:51 CET] <Johnjay> JEEB: any ideas?
[01:20:23 CET] <sfan5> ok libtls backend works so far
[01:20:50 CET] <sfan5> only problem is that libtls seems to force ocsp for tls servers which makes zero sense as it always produces an error (?)
[01:44:56 CET] <Johnjay> If I want to stream audio from windows should I use obs?
[01:48:44 CET] <geri> obs seems good
[01:48:50 CET] <geri> vlc player maybe too
[01:49:12 CET] <geri> https://www.videolan.org/vlc/streaming.html
[02:26:11 CET] <redrabbit> obs is easy to setup
[03:56:58 CET] <zyme> may I ask a couple newbie ffmpeg questions?
[04:08:13 CET] <zyme> two things mainly, to utilize any hardware acceleration, do they need to be specified in the command syntax? (like Cuda or any(if any) advantage from my Intel HD 4000 (Ivy Bridge) - Optimus Laptop setup)? Also from the URL: http://ffmpeg.org/ffmpeg.html#Detailed-description the simple little diagrams and explanations make me wonder, instead of transcoding or decoding or pure-encoding, is it possible to partially decode a
[04:08:13 CET] <zyme> file or video stream to a simpler+Somewhat Larger video encoded format (without a full decoding/re-encode) then set the output detailed specifications to detail the vid type as the most complicated encoded mode left from FFmpeg's output? (as if it were that video type instead already)?
[04:13:04 CET] <c_14> first question, yes
[04:13:12 CET] <c_14> second question, no idea what you're asking
[04:20:21 CET] <zyme> Say you have a newer version of an encoding than your device can decode, I assume the decoded frames are some form of basic MPEG after the decoding is finished before the encoding proccess starts - completing a transcode, say in the procces of H.263/divx/xvid to x264 (or its reverse) - am I right so far?
[04:20:52 CET] <c_14> no
[04:21:01 CET] <c_14> the decoded video is raw video frames
[04:21:15 CET] <c_14> just direct pixel data
[04:25:59 CET] <zyme> okay, then depending on the next answer I might have one more question; if you have a newer+more advanced form of the same codec encoding a video, such that the newer encoding of the format isn't backwards compatible with the decoder your trying to use, is it possible to partially decode a codec without the full raw decode and recompression ?
[04:26:39 CET] <c_14> nope
[04:29:01 CET] <zyme> so then the only way to transcode is to make the video completely raw and then reencode - and use a lossless format to encode that if you need it to not loose detail?
[04:29:20 CET] <c_14> pretty much
[04:32:13 CET] <zyme> how much of a difference is there in cpu power required for encoding h.263 vs h.264? (which is better for speed if size/bandwith isn't  an issue) and do they have lossless encoding modes?
[04:32:41 CET] <c_14> using x264 with ultrafast is probably the best way to go
[04:32:46 CET] <c_14> and x264 has a lossless mode
[04:34:12 CET] <zyme> my slowest network bottleneck is 300mbit/s but it could be upgraded to USB2-limited gigabit ethernet throughput.
[04:42:42 CET] <c_14> 300mbit is plenty
[04:42:48 CET] <c_14> (for most cases)
[04:43:03 CET] <aphirst> 32K of RAM etc
[04:43:24 CET] <geri> x264 ultrafast?
[04:43:37 CET] <geri> can you call it from opencv?
[05:17:26 CET] <_Vi> Is the fact that "-to" option, unlike "-t", is applicable only to output file, not input file deliberate or accidental? Is a patch that makes `ffmpeg -ss 1:20 -to 1:25 -i ...` work welcome?
[05:18:30 CET] <_Vi> The change looks trivial.
[07:00:59 CET] <hiihiii> dammit
[07:01:20 CET] <hiihiii> I've spent the night trying to convert some files on a server
[07:02:15 CET] <hiihiii> now that I've finished; I have audio desync at the end of  files
[07:12:40 CET] <hiihiii> I have VRF and CFR issues
[07:13:15 CET] <hiihiii> my inputs are CRF 25fps
[07:13:52 CET] <hiihiii> when I concat them in an FLV container I get CRF 25fps but no audio
[07:14:26 CET] <hiihiii> then when I concat them in MP4 I get audio but VRF 30fps
[07:16:08 CET] <blap> I used to have VCR issues.
[07:17:39 CET] <hiihiii> I'll give it another shot, this time with vsync
[08:04:41 CET] <twid> what is AVBuffer and what is it used for? is it some kind of memory pool for AVPacket/AVFrame?
[09:12:40 CET] <twid> what is AVBufferPool?
[10:14:57 CET] <rIMpossible> [6~[6~Good morning
[10:50:07 CET] <styler2go> Any idea why i get this error? Could not find tag for codec pcm_alaw in stream #1, codec not currently supported in container. command: ffmpeg -i rtsp://192.168.178.75:554/11 -acodec copy -vcodec copy -f mp4 -y test.mp4
[10:51:52 CET] <sfan5> mp4 does not support storing pcm_alaw audio, just like the message says
[10:52:54 CET] <sfan5> try with -acodec pcm_s16le
[10:55:47 CET] <styler2go> this works well.. and i thought it can't detect the rtsp stream correctly
[11:03:29 CET] <mbrr> Has anyone been used ffplay on mobile? It's the only thing that will play my stream with low latency but I can only run it on my laptop.
[11:03:53 CET] <mbrr> -been*
[11:06:29 CET] <mbrr> Multicast audio only. opus in mpegts over udp.
[11:13:26 CET] <mbrr> Successfully cross-compiled ffmpeg (with ffplay) for raspberry pi. I think most mobile devices use the same chip architecture as the pi. Is there a way to bring apps from the pi (raspbian) to android?
[11:14:01 CET] <BtbN> no
[11:14:13 CET] <BtbN> ffplay is only a proof of concept, no proper player
[11:18:30 CET] <mbrr> okay. thanks BtbN.
[11:19:00 CET] <mbrr> know of a proper player that plays network streams with little or no buffer?
[11:19:57 CET] <BtbN> Every normal player is optimized for smooth playback, and not zero latency
[11:20:31 CET] <mbrr> Just need something close to "fflpay -fflags nobuffer udp:/..." that works on mobile
[11:32:56 CET] <mbrr> Maybe I'll get ffplay in android if I can bring over ffmpeg and SDL.
[11:35:53 CET] <mbrr> Or maybe I should start from scratch and make a new player. It only needs to demux and play one specific codec/container/frequency/etc.
[11:40:12 CET] <Mavrik> That sounds like a hugely wrong way to do ti.
[11:40:25 CET] <Mavrik> Especially the bringing over SDL part, Android UI toolkit isn't really desktop linux compatible.
[11:40:27 CET] <utack> vlc on android?
[11:40:43 CET] <Mavrik> Not to mention lack of HW decoding murdering the battery (and performance)
[11:43:45 CET] <mbrr> can't get vlc to play the stream. not on linux or android. Someone suggested building vlc from the latest source but I haven't been able to make that happen yet.
[11:44:09 CET] <Mavrik> And what is the stream?
[11:44:42 CET] <mbrr> opus audio. mpeg-ts container.
[11:45:31 CET] <mbrr> grabbing audio from mic
[11:46:58 CET] <MoziM> how do programs like adobe premiere go backwards in a video frame by frame?
[11:55:12 CET] <mbrr> last resort is to start over trying to figure something out with WebRTC. Hard to let go of my current progress though because the server side is fine which I thought would have been the hard part.
[11:58:24 CET] <MoziM> what's the problem?
[12:00:20 CET] <mbrr> playing a low latency network stream on mobile. currently only plays on my laptop with ffplay and nobuffer flag.
[12:00:32 CET] <Mavrik> mbrr, I don't understand why do you think changing to a different wire format would solve anything
[12:01:08 CET] <Mavrik> Opus is supported on Android 5.0+ out of the box, probably the player you're using messes something up or you're muxing it into something unexpected
[12:02:59 CET] <mbrr> "ffmpeg -ac 1 -f alsa -i hw:1 -codec:a libopus -b:a 32k -application lowdelay -flags low_delay -f mpegts udp://239.255.0.1:5004"
[12:12:53 CET] <mbrr> or -f rtp_mpegts rtp://239.255.0.1:5004 seems to get closer to vlc playing it. vlc will show the media info> publish = FFmpeg with this protocol
[14:13:04 CET] <styler2go> what's a quick transcoding? i have the problem that if i use copy, one minute is 31mb big, if i do h264 native to libx264 it's only about 2mb but it does drop frames
[14:13:33 CET] <JEEB> libx264 shouldn't drop frames at all
[14:13:45 CET] <JEEB> so the dropping is happening somewhere else in the chain
[14:13:54 CET] <JEEB> libx264 as an encoder gets image in, and then outputs an image
[14:16:09 CET] <styler2go> hmm
[14:16:15 CET] <styler2go> now libx264 is also too big
[14:18:17 CET] <BtbN> You cannot shrink a video indefinitely, unless you are fine with it being a smudge pile of pixels beyond recognition
[14:18:57 CET] <styler2go> I am trying to record a security camera stream. So if one minute would be 31mb it would take me more than a TB to save one month
[14:19:23 CET] <styler2go> besides it's only 720p20 and 6144k bitrate
[14:19:29 CET] <BtbN> Sounds about right?
[14:19:53 CET] <styler2go> 1 minute = 31mb sounds really huge for "just a security camera"
[14:20:20 CET] <BtbN> Sounds on point for a 720p stream with ok quality
[14:20:59 CET] <JEEB> styler2go: do note that you pointed at absolutely no parameters that you used if you set any at all.
[14:21:10 CET] <JEEB> you can make libx264 slow as hell and you can get GOP length to longer
[14:21:15 CET] <JEEB> *set GOP length
[14:21:35 CET] <JEEB> and you can also tweak how much bits the encoder is ready to use
[14:22:35 CET] <styler2go> JEEB: i did some tests, i tried copy for video and i tried to let ffmpeg decide on its own
[14:22:44 CET] <JEEB> lol
[14:22:50 CET] <JEEB> ffmpeg doesn't decide jack shit
[14:22:52 CET] <DHE> ffmpeg doesn't do that
[14:23:01 CET] <JEEB> with libx264 you get x264's defaults, that is all
[14:23:12 CET] <JEEB> CRF 23 and preset medium and most likely something like a GOP of 250 frames
[14:23:14 CET] <DHE> file extensions determine formats. formats determine preferred codecs. and codecs have defaults
[14:23:18 CET] <styler2go> i mean i let ffmepg decide which codec to use
[14:23:47 CET] <BtbN> The defaults are usually bad, don't use them
[14:23:59 CET] <JEEB> well, x264 has somewhat better ones because people complained :D
[14:24:09 CET] <JEEB> FFmpeg used to have the crazy defaults before
[14:24:10 CET] <styler2go> https://p.styler2go.de/7585349 that's an example output
[14:24:14 CET] <DHE> you really are best off explicitly telling ffmpeg what to do. it's not like it's checking the resolution and framerate to make an educated guess of what the quality is
[14:24:21 CET] <BtbN> Yeah, crf23 with medium preset is actually a sensible default
[14:24:44 CET] <JEEB> because the global defaults are something like 200kbps and MPEG-1 features enabled or so
[14:24:50 CET] <JEEB> which was "u w0t m8t" level
[14:25:12 CET] <JEEB> (and you still get this with most lossy encoders in FFmpeg
[14:25:29 CET] <styler2go> the "default" setting creates a 6mb file for 1 minute which doesn't really look different to my 31mb version
[14:25:55 CET] <JEEB> oh, you're even capturing :P
[14:26:02 CET] <JEEB> for capture I deeply recommend capturing as-is if possible
[14:26:09 CET] <JEEB> because you don't want to drop crap
[14:26:20 CET] <JEEB> then you can separately encode those for archival
[14:26:22 CET] <styler2go> so, vcodec copy?
[14:26:27 CET] <JEEB> just -c copy
[14:26:43 CET] <JEEB> and I recomend using something that isn't mp4 for the capture in case your system derps
[14:26:46 CET] <styler2go> it then tells me that mp4 can't handle the audio codec
[14:26:53 CET] <JEEB> use either mkv or ts
[14:27:06 CET] <styler2go> ts seems to work out
[14:27:09 CET] <styler2go> what's ts?
[14:27:12 CET] <JEEB> MPEG-TS
[14:27:14 CET] <JEEB> transport streams
[14:27:24 CET] <JEEB> just make sure the audio plays after muxing
[14:27:51 CET] <JEEB> styler2go: MPEG-TS isn't meant for seeking but it doesn't have any indexes so for the first pass of streaming capture it makes sense
[14:28:07 CET] <JEEB> even if your PC crashes you should have some data :P
[14:28:09 CET] <styler2go> using tsa it ended up on a 33mb file
[14:28:58 CET] <JEEB> of course, I mean it will not change the video. I just don't see the reason for re-encoding when capturing this. you want to capture in pieces and then re-encode for archival when you don't have a realtime requirement
[14:30:58 CET] <styler2go> i don't have that much storage though
[14:34:34 CET] <JEEB> and realtime with a low-end CPU will end up a problem anyways
[14:35:10 CET] <styler2go> I am not sure which cpu is in my media pc to be honest but yes, probably low-end cpu
[14:35:22 CET] <styler2go> it has a dedicated gpu though
[14:35:22 CET] <JEEB> whether or not you can compress that well enough in real time depends on how compressed it is already
[14:35:44 CET] <JEEB> the dedicated hardware for video decoding and encoding on the GPUs are not made for compression
[14:35:57 CET] <JEEB> so you will not save bits
[14:47:22 CET] <styler2go> https://p.styler2go.de/6011721 that's some good quality haha
[14:47:54 CET] <styler2go> (i'm just messing aroudn with quality/file size to find a good "inbetween" for me
[14:50:02 CET] <Sbur3> Looking to recover a video file that I had been able to watch in the past.  Now, all the media players I have tell me that there is something wrong with the file
[14:50:10 CET] <Sbur3> Anyone wanna help?
[14:50:31 CET] <styler2go> Sbur3: kann ffprobe read the file codec etc?
[14:50:57 CET] <Sbur3> styler2go: Can you walk me through ffprobe?  An app or a command line?
[14:52:38 CET] <styler2go> it's a command line tool coming with ffmpeg
[14:52:56 CET] <styler2go> ffprobe -i "your file"
[14:54:52 CET] <Sbur3> styler2go: /home/steve/Desktop/262706470.mp4: Invalid data found when processing input
[14:55:25 CET] <styler2go> Doesn't sound too good.. Maybe someone else in here can help :/
[14:55:54 CET] <Sbur3> styler2go: Ok, but WinFF didn't figure out how to convert it
[14:56:27 CET] <Sbur3> styler2go: This stuff preceeds the other line ...
[14:56:34 CET] <Sbur3> styler2go: [mov,mp4,m4a,3gp,3g2,mj2 @ 0x5593048210c0] Format mov,mp4,m4a,3gp,3g2,mj2 detected only with low score of 1, misdetection possible!
[14:56:35 CET] <Sbur3> [mov,mp4,m4a,3gp,3g2,mj2 @ 0x5593048210c0] moov atom not found
[14:57:43 CET] <Sbur3> styler2go: And what bothers me is that it was possible to watch before ...
[14:57:57 CET] <styler2go> maybe your drive is corrupted?
[15:00:22 CET] <Sbur3> styler2go: Nope, cause other video files work
[15:00:41 CET] <styler2go> hmm i can't help you with that, sorry. maybe someone else has an idea
[15:00:47 CET] <Sbur3> styler2go: Recovered that file and another off a cd
[15:01:25 CET] <Sbur3> styler2go: Thanks for having tried to help.  It is appreciated
[17:29:26 CET] <AnonBaiter> hi
[17:29:53 CET] <AnonBaiter> if I recall correctly, there is a way to add a header into some headerless video/audio data through ffmpeg
[17:29:58 CET] <AnonBaiter> is there a way to do this?
[17:34:25 CET] <DHE> possibly, but it will involve regenerating the file.
[17:34:41 CET] <DHE> ffmpeg -i input.ext -c copy [options to insert your desired metadata here] output.ext
[17:39:32 CET] <AnonBaiter> "-c copy"?
[17:39:54 CET] <AnonBaiter> and what about metadata? do I have to insert some kind of command such as framerate, resolution, colorspace, etc.?
[17:43:58 CET] <DHE> well you never specified what metadata you wanted
[17:45:34 CET] <AnonBaiter> all I know is that the video resolution is at least 256x256, and it contains MPEG2 data
[18:16:24 CET] <styler2go> is it possible to create a timelaps with ffmpeg? all files in a folder to one video with 10 times playspeed?
[18:17:11 CET] <geri> do you have images?
[18:17:16 CET] <AnonBaiter> ffmpeg.exe -i "00010257_0x01620000.bin" -f mpeg2video -s 256x256 -vframes 30 -pix_fmt yuv420p "00010257_0x01620000.m2v"
[18:17:19 CET] <AnonBaiter> kinda like this...
[18:17:28 CET] <styler2go> no, video files
[18:31:29 CET] <styler2go> https://p.styler2go.de/927006 Why is the filename unsafe?
[18:34:05 CET] <sfan5> read this http://ffmpeg.org/ffmpeg-formats.html#Syntax
[18:35:02 CET] <geri> sfan5: use a timer. When image is not ready write last available image?
[18:35:13 CET] <geri> wrong chan
[18:37:12 CET] <styler2go> sfan5 i am already escaping it.. what else should i do
[19:00:26 CET] <sfan5> styler2go: you need the "ffconcat version 1.0" to allow unsafe filenames
[19:05:33 CET] <styler2go> what's an unsafe filename anyway?
[19:13:54 CET] <klaxa> maybe something like ../somefile ?
[20:19:50 CET] <laduke-13> can I pipe an image into ffmpeg and have it loop forever so I can pipe that into a player?
[20:20:42 CET] <Toba> https://superuser.com/questions/1041816/combine-one-image-one-audio-file-to-make-one-video-using-ffmpeg
[20:20:47 CET] <Toba> ^ this should get you in the right direction
[20:22:30 CET] <laduke-13> thanks! i've been able to output to a file of a specific length so far
[20:23:13 CET] <Toba> if you set it to loop it should run as long as you want it to
[20:40:24 CET] <arpu> hi how can i crop a video without reecode and using stream copy?
[20:40:39 CET] <arpu> filert:v  and crop only works with reencode :/
[20:40:45 CET] <arpu> any hint?
[20:41:26 CET] <c_14> depending on the container you can try container cropping
[20:45:46 CET] <arpu> c_14, container is flv (rtmp h264)  or webm (icecast vp9)
[20:46:43 CET] <furq> only mkv supports that afaik
[20:46:46 CET] <furq> so maybe webm does, idk
[20:46:46 CET] <c_14> webm should do it since it's a subset of matroska and matroska has it. Question is if players support it
[20:46:56 CET] <furq> also yeah that's the bigger problem
[20:47:09 CET] <kepstin> well, webm is a *subset* of matroska, they might not have included support for that :)
[20:47:19 CET] <furq> right
[20:48:03 CET] <arpu> i think i can use mkv with icecast
[20:50:19 CET] <arpu> hm but if i use the mkv container cropping  the filesize would be the same? as the original?
[20:50:48 CET] <c_14> yes
[20:50:55 CET] <JEEB> > container cropping
[20:51:08 CET] <JEEB> that  is one of the worst features in matroska
[20:51:23 CET] <JEEB> just because it was never really specified in the way that the author intended
[20:51:41 CET] <arpu> so why it is not possible to delete some pixels without reencode?
[20:51:53 CET] <JEEB> it was meant so that the demuxer could "fix" broken AVC headers and rewrite them in the demuxer
[20:52:01 CET] <JEEB> and thus the amount of was limited
[20:52:15 CET] <JEEB> but taht wasn't specified and then you got 1920x1080 clips with enough cropping for 1440x1080
[20:52:24 CET] <JEEB> which of course then the original implementations didn't support
[20:52:40 CET] <JEEB> unfortunately both mpv and vlc seem to support it, so if that is what you want to utilize it should work there :P
[20:53:00 CET] <arpu> for me the idea is to have a 3d stereo live video and on normal desktop i will show the mono video so the half height of the video
[20:53:01 CET] <JEEB> arpu: the proper way would be to rewrite the headers, I think someone was making a bit stream filter for that for MPEG-2 Video, AVC and HEVC
[20:53:27 CET] <JEEB> arpu: so how are you supposed to get the full picture from the matroska then :P
[20:53:28 CET] <arpu> JEEB,  yes but this does not save me 50% of the bandwidth right?
[20:53:51 CET] <JEEB> if you tell in the matroska cropping tag that "always remove this amount of crap"
[20:58:08 CET] <arpu> so this could work?
[20:58:27 CET] <JEEB> if you limit yourself to specific players
[20:58:39 CET] <JEEB> and how the hell are you saving bandwidth?
[20:59:02 CET] <JEEB> you're still transferring the same stereo live video, but just telling the player to not show half of it
[20:59:57 CET] <arpu> ok this is the point
[21:00:27 CET] <arpu> i will playing the video on web with dash
[21:00:57 CET] <arpu> the idea was  to have the half size of the video without reencoding ..
[21:01:06 CET] <arpu> but yeahh this is not possible
[21:01:07 CET] <JEEB> bandwidth will be exactly the same
[21:01:21 CET] <JEEB> with MVC you could have the second view as separate and dynamically mux it in
[21:01:44 CET] <JEEB> because you can have references to the normal pictures but otherwise the MVC pictures are completely separate
[21:01:53 CET] <JEEB> but no idea if that is *half* the bit rate
[21:02:02 CET] <JEEB> since I've not seen how effective MVC coders are
[21:02:20 CET] <arpu> what is MVC?
[21:04:15 CET] <JEEB> multi-view coding
[21:04:21 CET] <JEEB> a feature extension to AVC
[21:04:27 CET] <JEEB> which was not widely implemented
[21:04:54 CET] <JEEB> you have the base layer for the first view, and then you have packets for the MVC stuff
[21:04:59 CET] <JEEB> which a normal AVC decoder can ignore
[21:06:14 CET] <arpu> ok thx for  all the information!
[21:06:56 CET] <JEEB> it would just be o9k times simpler if you just had the multiview thing completely separate
[21:07:15 CET] <JEEB> sure you will need two encoder instances but you would not have to care about various bullshit
[21:07:49 CET] <arpu> yes this is a problem  to reencode every live video
[21:08:09 CET] <arpu> and 3d stereo are 4k or 8k
[21:13:13 CET] <mbrr> I have MPV open in android-studio. Is there any chance I'll be able to tweak it to not buffer or cache a network stream?
[21:16:34 CET] <JEEB> mbrr: probably #mpv-android is more relevant
[21:17:12 CET] <mbrr> that is what I meant
[21:17:33 CET] <JEEB> yes, but that has its own channel, this is #ffmpeg
[21:17:46 CET] <mbrr> ooh. k. thanks.
[22:36:24 CET] <Cracki> you're talking stereo? dropping the second eye gives very little bitrate savings. the two pictures are similar, so the "double picture" costs little to encode
[22:39:46 CET] <Cracki> a cropping bitstream filter might be _possible_, but it'll have to reencode blocks that refer to cropped away pixels
[00:00:00 CET] --- Sun Nov  5 2017


More information about the Ffmpeg-devel-irc mailing list